Last summer, Josh Simons wrote a blog entry titled Summer of RDMA, summarizing work done by our intern, Adit Ranadive, who looked at RDMA performance on ESXi using VM DirectPath I/O with QDR InfiniBand cards passed through to VMs.
At the same time, I had been exploring how to exploit RDMA for accelerating ESXi hypervisor services like vMotion, and some ideas around how to enable RDMA within a guest while still maintaining the ability to vMotion such VMs.
In March 2012, Josh Simons and I delivered a talk about RDMA on vSphere: Status and Future Directions. In the talk, I discussed several different altenatives to expose RDMA capabilities to vSphere guests, including an "Option F" which I refered to as the "Holy Grail" of the different approaches.
This summer we were thrilled to have Adit back in the Office of the CTO for another internship, this time specifically prototyping the "Holy Grail," now officially called "vRDMA," for a paravirtual RDMA device.
Before Adit heads back to school to complete his Ph.D. at Georgia Tech, I had the opportunity to sit down with him and interview him about his internship with us this summer. Check out the video here, to get a glimpse of one of the many awesome internship projects we hosted this summer at VMware, chock full of bright and talented students from some of the best schools from around the world.
We've made rapid progress in implementing the prototype and are close to getting it working. Once it's working and we can run the OpenFabrics "perf-tests" on vRDMA, I look forward to sharing our preliminary performance results of using this approach to offer RDMA services to vSphere virtual machines.